A 12-month 2-wave prospective longitudinal study was conducted among junior middle school students in Guangzhou and Chengdu, China; baseline and follow-up surveys were performed in December 2018 and 2019, respectively. The inclusion criteria at baseline included full-time grade 7 and grade 8 students, aged 12‐14 years, who had played internet games in the past 12 months at baseline, and were willing to participate in this survey. The selection of Grade 7 and 8 students was driven by their behavioral risks and environmental shifts specific to early adolescence. Behaviorally, this period involves heightened sensation-seeking and emotional instability, making the immediate gratification of gaming highly attractive [52]. Environmentally, these students are navigating the stressful transition to secondary education and experiencing reduced parental supervision regarding digital device use compared to primary school. Those Grade 9 students were excluded as they would leave school after 1 year and could not be followed up.

With the assistance of local education sectors, 4 and 2 junior middle schools were conveniently selected, and all grade 7 and 8 students of the selected schools were invited for participation; the in-school surveys were approved by school principals. The anonymous survey was self-administered by the students in the classroom setting in the absence of schoolteachers. Before the survey, well-trained field workers prebriefed the students on the objectives, content, logistics, and voluntary nature of this study, and that submitting a completed questionnaire implied informed consent for participation. Written informed consent was not collected to maintain anonymity. Such information was also printed on the cover page of the structured questionnaire. In addition, field workers assisted the students (eg, answering inquiries) and did the quality check during and after the survey, respectively. After 12 months, the students completed the same questionnaire with the same data collection procedures. The reporting of this study follows the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines [53], and the checklist is included in the supplementary material.

At baseline, 2463 completed the questionnaire, and a total of 77.4% (1906/2463) questionnaires were matched for both baseline (T1) and follow-up (T2) surveys. The lost to follow-up (530/1906, 21.8%) group was more likely than those followed-up to be older, Chengdu participants, male, not living with both parents, having father’s and mother’s educational levels of junior middle school or below, and having higher scores of IGD and some domains of perceived immediate and long-term benefits and losses (Table S1 in Multimedia Appendix 1). Among the matched questionnaires, 25.9% (494/1906) were excluded due to not having played internet games in the past 12 months at T1, 5.4% (103/1906) due to missing data in key variables (eg, gaming behaviors and IGD), and 7.1% (136/1906) due to age below12 or above 14 years. The final sample size was 1173 in this study.

Descriptive statistics and normality tests (via Kolmogorov-Smirnov and Shapiro-Wilk) were performed. Little’s Missing Completely at Random (MCAR) test was conducted to evaluate the pattern of missing data; Multiple Imputation was used to handle missing data. Paired-sample t tests were used to test for within-individual differences in levels of perceived immediate and long-term benefits and losses at both T1 and T2. Such data analysis was performed by using SPSS software (version 26.0; IBM Corp). A 2-sided P value <.05 was defined as statistically different.

Cross-lagged panel analyses with the maximum likelihood robust estimator were performed using Mplus 8.3; the analyses were conducted across 20 imputed datasets, and parameters were pooled according to Rubin’s rules. In total, 9 models were fit to examine the potential bidirectional associations between each pair of perceived immediate versus long-term benefits and losses and IGD. These models had 3 types of correlations: (1) cross-sectional correlations between variables at the same time point, (2) autocorrelations between the level of the same variable at both T1 and T2, and (3) cross-lagged correlations between the variable at T1 and the other variables at T2. To better reveal the cross-lagged correlations (eg, perceived short-term benefits of mental health on IGD), both cross-sectional correlations and autocorrelations, as well as background factors (ie, age, sex, city of study, whether living with both parents, parental educational levels, and perceived family financial status) at T1 were controlled. Satisfactory model fit indices of the models included CFI≥0.90, TLI≥0.90, and SRMR≤0.08 [55].

This study was approved by the Survey and Behavioral Research Ethics Committee of the Chinese University of Hong Kong (number SBRE-18‐430). Participation in this study was entirely voluntary. Prior to data collection, all participants were informed of the study objectives, procedures, and their rights as research participants. Given the anonymous nature of the survey, written informed consent was not obtained. Instead, informed consent was implied by the voluntary completion and submission of the questionnaire. Such information was also printed on the cover page of the structured questionnaire. All data were anonymous and no direct personal identifiers were collected. To enable longitudinal matching while preserving anonymity, instead, the last 4 digits of the father’s and mother’s phone numbers and the last phonetic letters of the father’s and mother’s given names were collected for matching purposes. No identifiable information about participants is hence possible in any part of this study or supplementary material. No financial or material incentives were provided to participants for their involvement in this study. In addition, parental approval was sought, and the parental opt-out procedure was exercised.

Among all participants, the mean age was 12.5 (SD 0.6; range=12-14) years; 72.1% (846/1173) were from Guangzhou and 27.9% (327/1173) from Chengdu; over half (693/1173, 59.1%) were male; 15.6% (183/1173) did not live with both parents; 29.9% (350/1173) perceived that their family financial level was poor or very poor; nearly one-third of the participants’ fathers (361/1173, 30.8%) or mothers (335/1173, 28.6%) had received college or above education (Table 1). The mean IGD scores were 2.1 (SD 2.1; range=0-9) and 1.7 (SD 2.1; range=0-9) at T1 and T2, respectively.

This study was novel to investigate the levels of perceived immediate and long-term benefits in 3 dimensions (mental health, social relationships, and personal achievement) and perceived immediate and long-term losses in 6 dimensions (mental health, sleep quality, academic performance, family relationships, social relationships, and personal achievement). All these levels were comparably low (within a range of 0‐4); the mean scores for perceived benefits ranged from 1.1 to 1.5, while those for perceived losses ranged from 0.7 to 1.4. These findings reflect the nature of the general adolescent sample, where the majority of participants engage in gaming recreationally without reporting extreme levels of both positive and negative experiences. Nonetheless, the observations confirmed a previous study reporting comparable levels of perceived immediate versus long-term benefits [22] and further underscore the importance of integrating the time perspective into perceived benefits and losses. As the first study to investigate perceived losses, our findings suggest that adolescents may perceive both benefits and losses equally, and perceived losses should be a nonnegligible target in future interventions; future studies in this regard are greatly warranted for confirmation and elaboration.

This longitudinal study was also the first to reveal the interesting bidirectional relationships between perceived immediate and long-term benefits and IGD. Although existing cross-sectional studies reported that perceived benefits of improved mental health, social relationships, and personal achievement were positively associated with IGD [21-26], this longitudinal study only partially supports these findings and revealed the differential predictive effects of these perceptions on IGD. The significant results highlight the importance of perceived immediate mental health and personal achievement benefits. Plausibly, the immediate mental health benefits could offer quick and easily accessible relief from negative emotions for gamers. Furthermore, the instant gratification model postulates that behaviors (eg, internet gaming) providing immediate emotional relief are more likely to be repeated [56], potentially contributing to increased IGD risk. On the other hand, the self-determination theory postulates that satisfaction with the psychological need for competence is powerful in motivating behavior [57]. Accordingly, the sense of accomplishment gained from internet gaming could supply immediate feedback and rewards (ie, immediate personal achievement benefits), motivating repetitive gaming, thereby increasing IGD risk. In addition, the significant predictive effects involving immediate but not long-term benefits corroborate previous publications emphasizing immediate rewards over delayed gratification in the context of IGD [58]. Speculatively, long-term benefits on mental health and personal achievement from internet gaming might occur gradually and remain less tangible to adolescent gamers, leading to the nonsignificant results. The nonsignificant predictive effects regarding perceived social relationships benefits might be associated with the complexity of social interactions, which can be context-dependent and influenced by external factors outside gaming (eg, peer relationships and behavioral inhibition) [59,60]. It is speculated that social interactions in gaming might not necessarily translate into meaningful or fulfilling relationships, which might reduce their impacts on IGD development.

Conversely, IGD at T1 significantly predicted perceived immediate and long-term benefits of social relationships and personal achievement at T2, but not those of mental health. It suggests that those with more IGD symptoms might amplify perceived benefits related to social relationships and personal achievement. Internet gaming is often designed with structured opportunities for social interactions (eg, multiplayer engagement) and measurable achievements (eg, leveling up) [61]; these aspects might be more pronounced among those with more IGD symptoms, leading to stronger perceptions of these benefits over time. In addition, those with more IGD symptoms may prioritize social and achievement benefits from internet gaming over those from other sources over time [21,62]. In contrast, individuals with more IGD symptoms might experience negative mental health outcomes, such as anxiety and depression [5,8]. These negative psychological conditions may overshadow the mental health benefits and lead to the nonsignificant predictive effect of IGD on perceived mental health benefits.

Unexpectedly, all predictive effects of perceived immediate and long-term losses in 6 domains at T1 on IGD at T2 were statistically nonsignificant. While perceived benefits and losses might be perceived at comparable levels, the psychological impacts of immediate benefits might have stronger influences on developing IGD due to their ability to fulfill psychological needs like competence as aforementioned [20]. This could overshadow the importance of perceived losses in terms of affecting IGD. This speculation was also supported by the concept of reinforcement that positive reinforcement (eg, perceived benefits) tends to have more immediate and stronger effects on motivating behavior in comparison with negative reinforcement or punishment (eg, perceived losses) [63]. Nonetheless, such speculations should be tested in future studies.

Reversely, this study observed differential predictive effects of IGD at T1 on the 6 domains of perceived immediate and long-term losses at T2. First, IGD significantly predicted perceived immediate and long-term losses in mental health, sleep quality, academic performance, and family relationships, suggesting that IGD could lead to noticeable harm in these domains. These findings align with empirical evidence that IGD cases face common negative consequences, such as depression, disturbed sleep patterns, worsened academic performance, and strained family relationships [27,33,41]. Notably, these domains of perceived losses might be more directly observable or harder to rationalize over time and are recommended to be emphasized in IGD intervention programs. Second, IGD at T1 significantly predicted perceived long-term, but not immediate, losses in social relationships, indicating a potential lag effect. It is speculated that, initially, the social benefits of internet gaming might buffer against perceived immediate losses, but, over time, those with more IGD symptoms might experience weakened offline social ties and relationships [64], making the long-term losses more apparent and significant. Last, the predictive effects of IGD at T1 on perceived immediate and long-term losses in personal achievement were statistically nonsignificant, suggesting that individuals with more IGD symptoms might not associate their gaming habits with detriments in personal achievement. Plausibly, as aforementioned, internet gaming may fulfill one’s psychological needs of competence [65] and provide satisfactory substitutes for real-life achievements [66], decreasing perceived losses.

The current findings have practical implications, as gaming-specific cognitions are important modifiable constructs in IGD interventions (eg, cognitive-behavioral therapy) [67]. IGD prevention interventions may focus more on the immediate benefits on mental health and personal achievement. For instance, alternatives for immediate rewards of mental health and personal achievements (eg, outdoor physical activities) should be provided for adolescents to reduce their reliance on and sensitivity toward these gaming rewards. It might also be helpful to separate gaming achievement from real-world achievement and then educate adolescents on how to fulfill the latter [68]. In contrast, IGD treatment intervention may underscore the cognitions of potential losses in mental health, sleep quality, academic performance, and family relationships, as well as the long-term losses in social relationships, while perceived rewards of social relationships and personal achievement are recommended to be downplayed.

There are several limitations of this study. First, the attrition rate of this study was about 20%. Attrition analysis revealed that those lost to follow-up demonstrated stronger perceived benefit and losses in some domains, besides specific background differences (eg, age and sex). In addition, Little’s MCAR test showed that the missing data were not completely random. It suggests that the study sample may have stronger perceived benefits and losses than the baseline population. Nonetheless, multiple imputation was used in the cross-lagged panel analyses to address the missing data patterns, increasing the robustness of the current findings. Furthermore, the study population was selected from 2 cities in China through convenience sampling, which may lead to selection bias. Cautions are needed when extrapolating the results to other regions or countries. Second, as participants were asked to self-administer the questionnaire, there might be reporting bias, including recall bias and social desirability bias. Third, although this study provides valuable insights on the bidirectional relationships between perceived immediate and long-term benefits and losses and IGD, it comprised only 2 time-points within a 12-month follow-up period. Future longitudinal studies with more time lags and longer study periods are needed to confirm the findings. In addition, while several predictive effects were statistically significant, their associated effect size was modest. It suggests that, despite statistical significance, these associations might have limited practical impacts on individuals. However, it is important to note that small effect sizes could still bear practical implications at the population level. Modest increases or reductions in each predictor might yield substantial benefits for the population, though future intervention studies are needed for confirmation. Last, there might be other potential losses due to internet gaming that were not included in this study (eg, losses in physical health and the financial domain).

This study was novel in integrating the time perspective into perceived benefits and losses related to internet gaming and investigating their bidirectional relationships with IGD. The results are interesting in that only perceived immediate benefits of mental health and personal achievement predicted IGD, while IGD predicted perceived immediate and long-term benefits on social relationships and personal achievement, as well as perceived immediate and long-term losses in all 6 domains (except personal achievement). These findings enhance the understanding of the relationships between gaming-specific cognitions and IGD and shed new insights for relevant studies, taking into consideration the time perspective. The differential bidirectional relationships are also important for the design of IGD prevention and treatment interventions. Specifically, relevant cognitive components should target the specific temporal bias adolescent gamers possess, rather than focusing merely on the general benefits and losses of internet gaming.